Two main mutational processes operate in the absence of DNA mismatch repair

The analysis of tumour genome sequences has demonstrated high rates of base substitution mutagenesis upon the inactivation of DNA mismatch repair (MMR), and the resulting somatic mutations in MMR deficient tumours appear to significantly enhance the response to immune therapy. A handful of different algorithmically derived base substitution mutation signatures have been attributed to MMR deficiency in tumour somatic mutation datasets. In contrast, mutation data obtained from whole genome sequences of isogenic wild type and MMR deficient cell lines in this study, as well as from published sources, show a more uniform experimental mutation spectrum of MMR deficiency. In order to resolve this discrepancy, we reanalysed mutation data from MMR deficient tumour whole exome and whole genome sequences. We derived two base substitution signatures using non-negative matrix factorisation, which together adequately describe mutagenesis in all tumour and cell line samples. The two new signatures broadly resemble COSMIC signatures 6 and 20, but perform better than existing COSMIC signatures at identifying MMR deficient tumours in mutation signature deconstruction. We show that the contribution of the two identified signatures, one of which is dominated by C to T mutations at CpG sites, is biased by the different sequence composition of the exome and the whole genome. We further show that the identity of the inactivated MMR gene, the tissue type, the mutational burden or the patient's age does not influence the mutation spectrum, but that a tendency for a greater contribution by the CpG mutational process is observed in tumours as compared to cultured cells. Our analysis suggest that two separable mutational processes operate in the genomes of MMR deficient cells.